DeepSeek Raises Funding at $45B Valuation Amid AI Competition

Chinese AI lab DeepSeek is in talks to raise its first funding round at a reported US$45 billion valuation, according to TechCrunch. The valuation marks a significant increase from an estimated US$20 billion valuation just weeks earlier, following attention drawn to the company's AI models for their lower training costs. Founder Liang Wenfeng controls nearly 90% of the company.

Fundraising Motivation and Investor Interest

DeepSeek decided to seek funding after rivals attempted to recruit researchers by offering company shares as incentive. The China Integrated Circuit Industry Investment Fund may lead the funding round, while Tencent and Alibaba are also reportedly in talks to participate.

Model Performance and Pricing

DeepSeek's V4 model is reported to perform on par with top models from OpenAI and Anthropic on benchmark tests, according to the company. The competitive advantage extends to pricing: V4-Pro costs US$1.74 per million input tokens, while V4-Flash costs approximately US$0.14 per million input tokens—significantly below pricing for comparable US models.

The lower cost structure stems from DeepSeek's compute-efficient architecture, including a mixture-of-experts (MoE) design that activates only part of the model for each task, reducing computing needs during inference. According to DeepSeek, V4-Pro uses 27% of the computing power and 10% of the memory required by V3.2.

China's AI Self-Reliance Strategy

DeepSeek's funding round occurs as China seeks to build a more self-reliant AI infrastructure in response to US export controls on advanced chips. V4 is DeepSeek's first model tuned for Chinese chips such as Huawei's Ascend series, though the company's technical report indicates Chinese chips handle inference while training may still depend primarily on Nvidia hardware.

DeepSeek also releases open-weight models—trained parameters that others can use, fine-tune, and deploy—broadening AI development beyond US dominance. This strategy emphasizes algorithmic efficiency over exclusive access to the most advanced US hardware, though China's AI push remains partially dependent on Nvidia hardware.